Creating an attractive urban greenspace in Vancouver: An exploratory analysis

by Matthew Yau

Section 1: Introduction

Imagine that one day, some time into the future, you decide to start a new job in the city hall. On your first day of work, you are given a new task. Specifically, you are asked to try and come up with a plan that can improve how attractive your city's greenspace is. How would you go about approaching this task?

Answering the question of how you can create an attractive urban greenspace would be difficult, but it would also be rewarding. Specifically, it may be important to answer this question because the knowlege gained from this question could be used to improve real world outcomes. For one, this knowledge could have applications at the municipal level. For example, the City Council of Vancouver aims to plant a greater number of trees in the city within the next decade, and knowledge of attractive greenspace creation could be used to inform the city council on how it could plant trees in an improved, more attractive manner. In addition, knowing how to create attractive greenspaces could also have other benefits. For example, it could amplify the positive effects that greenspaces have on the city. To elaborate, previous work has shown that urban forests produce positive economic and societal outcomes, such as improving property values and community mental health. However, it is also possible that, knowing how to create, and creating, attractive urban forests, could produce even greater positive economic and societal returns, such as raising property values even further, and raising community mental health to a greater extent.

Therefore, the present investigation aims to investigate how attractive urban greenspaces could be created. The investigation explores this question by focusing on urban forests in Vancouver: What sorts of trees in Vancouver create the most attractive greenspaces? To carry out this investigation, three general steps were followed: First, data for the analysis were obtained and prepared (Section 2). Specifically, data about trees was obtained from the Open Data Portal of Vancouver, which is a dataset that contains general information about a large number of trees in Vancouver. Furthermore, data related to greenspace attractiveness was obtained from My Health My Community Project, which is a community survey that assessed the health and well-being of communities across Vancouver, including the quality of the community's environment.

Second, the investigation examines, in an exploratory manner, how urban forests in Vancouver are related to community perceptions of greenspace attractiveness (Section 3). Specifically, the investigation focuses on asking three guiding questions: i) Are the number of trees in a neighbourhood related to community perceptions of greenspace attractiveness? ii) Is the size of trees related to greenspace attractiveness? iii) Is the diversity of tree species related to greenspace attractiveness? Third and finally, the findings of the investigation are discussed, and implications and directions for future work are outlined (Section 4).

Section 2: Data Preparation

This section highlights the steps that I took in obtaining and preparing the data.

Importing data

Data wrangling

Section 3: Analysis

To investigate how urban forests relate to greenspace attractiveness, the following three questions were examined:

To examine these questions, the following variables (presented in the tables below) were examined as potential independent variables that could infleunce greenspace attractiveness.

Independent variables
Variable Description Variable type Question
name Local area in which the tree is planted Categorical (nominal) Is the number of trees related to greenspace attractiveness?
diameter Diameter of the tree (in inches) Numeric Is the size of trees related to greenspace attractiveness?
height_range_id Height range of the tree, each unit represents 10 feet Categorical (ordinal) Is the size of trees related to greenspace attractiveness?
species_name Name of the species of tree Categorical (nominal) Is the diversity of tree types related to greenspace attractiveness?
genus_name Genus of the tree Categorical (nominal) Is the diversity of trees types related to greenspace attractiveness?
common_name Name of the tree Categorical (ordinal). Is the diversity of tree types related to greenspace attractiveness?
Dependent variables

To capture greenspace attractiveness, community perceptions of greenspace attractiveness from My Health My Community Project were used. Because there are a smaller number of data points within this greenspace data set, where each community contains only one aggregate measure of greenspace attractiveness, greenspace attractiveness was conceptualized as an ordinal variable consisting of 19 levels. Specifically, greenspace attractiveness was conceptualized as the rank order from the most attractive greenspace community to the least attractive community, where higher rank indicates greater greenspace attractiveness.

Variable Description
nat_surround_rating The percentage of residents within a neighbourhood who agreed that their surrounding greenspace is attractive Ordinal
Neighbourhood Natural surrounding rating score Rank of greenspace attractiveness
West Point Grey 96.7 1
Kitsilano 94.3 2
West End 93.3 3
Dunbar-Southlands 92.3 4
South Cambie/Oakridge 86.0 5
Fairview 83.3 6
Killarney 82.8 7
Downtown 81.9 8
Riley Park 80.4 9
Shaughnessy/Kerrisdale/Arbutus-Ridge 80 10
Mount Pleasant 75.5 11
Grandview-Woodland 74.8 12
Hastings-Sunrise 70.1 13
Kensington-Cedar Cottage 66.8 14
Renfrew-Collingwood 63.9 15
Victoria-Fraserview 63.6 16
Strathcona 63.2 17
Marpole 61.0 18
Sunset 56.0 19

Analytical approach

To generate insights, visualizations were used to examine whether the key independent variables tended to co-occur with the dependent variables. That is, do the most attractive greenspaces tend to contain more of an independent variable, such as a greater number of trees? Do less attractive greenspaces tend to contain less of an independent variable, such as less tree diversity?

Section 3.1: Preliminary analysis

General information about the dataset is provided below, such as the column types or summary statistics about the variables of interest.

Column types and missing values

Summary statistics

Is tree count related to greenspace attractiveness? A bar graph was used to answer this question, plotting the number of trees for each neighbourhood, with neighbourhoods being sorted from the most attractive greenspace to the least attractive. Examining this graph showed that tree count does not seem to be related to green space attractiveness.

3.3a) Boxplots of tree size across neighbourhoods

Is the height of trees related to greenspace attractiveness? To answer this question, boxplots were used to compare the size of trees between attractive and less attractive greenspaces, to enable comparison between central tendencies as well as variations. The same previous sorting order was used to distinguish between attractive and less attractive greenspaces.

Examination of this graph revealed that, the variation of tree sizes was similar across attractive and less attractive greenspaces. However, comparing the median reveals that, on average, attractive greenspaces tend to contain taller and wider trees, whereas less attractive neighbourhoods tend to contain shorter and thinner trees.

Why do attractive greenspaces have, on average, larger trees? Other visualizations may provide greater detail insights into this question, beyond what boxplots could capture. For example, although boxplots enable easy comparisons between central tendencies and variations, they are less effective in representing how the number of tall and wide (or short and thin) trees differ across neighbourhoods. Can we visualize how the number of trees (of varying sizes) differ across neighbourhoods?

3.3b) Heatmaps of tree size distribution across neighbourhoods

To examine how the number of different tree sizes differed across attractive and less attractive greenspaces, a heatmap was created, with color being used to represent the count of trees within every size interval, across neighbourhoods.

Examining this graph reveals that, the previous finding that attractive greenspaces have relatively greater tree sizes is mostly because less attractive greenspaces have a much larger amount of small trees, which drives down the average for these neighbourhoods.

The current heatmap is effective in showing how smaller trees differ across attractive and less attractive greenspaces, but it is less effective in comparing how larger trees differ across these greenspaces. This is because there are many more small trees, compared to large trees, in Vancouver. Since most of the color is used to represent a large number of small trees, this makes color less sensitive to variations in the number of large trees across neighbourhoods.

3.3c) Layered histogram of tree sizes across different neighbourhoods (filtered by legend)

To more effectively examine how large trees differs across attractive and less attractive greenspaces, a filterable histogram was created. This histogram enables the comparison of distributions between specified neighbourhoods (which is specified on legend), where histograms also capture the number of large trees.

Examination of this graph, by filtering for attractive greenspaces and less attractive greenspaces at the same time, reveals that, on average, more attractive greenspaces (e.g. West Point Grey, Kitsilano, Dunbar-Southlands) tend to contain a greater number of large trees compared to less attractive greenspaces (e.g. Sunset, Marpole, Strathcona).

Is the diversity of trees related to greenspace attractiveness? To examine this question, a filterable bubble plot was created, which can be filtered by different definitions of tree diversity, such as genus or species. The size of the bubble is sensitive to the count of the trees, and color is used to differentiate between species.

Examination of this graph revealed that the diversity of trees does not appear to differ across attractive and less attractive greenspaces.

Section 4: Discussion

Main findings

The present investigation aimed to gain a greater understanding of how attractive urban greenspaces could be created, by focusing on the urban forests of Vancouver and how they are related to community perceptions of greenspace attractiveness. Three potential questions were explored.

First, do the count of trees differ across attractive and less attractive greenspaces? A bar chart comparing the number of trees across attractive and less attractive greenspaces showed that the amount of trees is not related to perceptions of greenspace attractiveness. Second, do tree sizes influence greenspace attractiveness? A boxplot comparing central tendencies and variations showed that, on average, attractive greenspaces tended to contain larger trees (taller, wider) compared to less attractive greenspaces. Follow up analysis in the form of heatmaps showed that, the greater average size of trees in attractive greenspaces is mostly due to less attractive greenspaces containing a greater number of smaller trees. Furthermore, layered histograms filtering for the neighbourhoods of interest also showed that the higher average size of trees in attractive greenspaces is also partially, but to a small extent, due to attractive greenspaces containing a greater number of large trees. Third and finally, the investigation explored if tree diversity is related to greenspace attractiveness. A bubble plot sensitive to the number of different tree types showed that there does not seem to be a relationship between tree diversity and greenspace attractiveness.

Implications and future directions

All in all, the findings from the current investigation suggest two patterns of results. Specifically, increasing the number and diversity of trees within a neighbourhood may not be effective ways to improve greenspace attractiveness. Furthermore, increasing the number of larger (taller, wider) trees may be effective ways to improve greenspace attractiveness. Although it would be exciting to apply such insights to the real world, such as informing city officials how they could create attractive greenspaces, further work is needed before the claims of the current investigation could be applied to the real world. In this light, I highlight a few directions for future research.

First, future work should also aim to replicate the results of the current study with greater methodological rigor. One way that the methodology of the current investigation could be improved is to change the nature of the analytical approach. For example, to find that attractive greenspaces contain a larger number of large trees, the current study used layered bar charts to compare the count of large trees between attractive and less attractive greenspaces. However, such an approach would make it difficult to have comparisons of tree counts between more than two neighbourhoods at once. In turn, an inability to visualize the number of tall trees in all attractive and less attractive neighbourhoods could skew interpretation and lead to inaccurate conclusions. Therefore, one analytical improvement may be to compare the number of large trees in all neighbourhoods using stacked bar charts, rather than layered. An example of doing so has been provided below.

Another methodological improvement that can be made is to increase the sensitivity of the key measures of interest. For example, the current investigation used an ordinal variable as the key dependent variable of interest: that is, the rank order of greenspace attractiveness. However, more effective and sensitive measures could be used. For example, residents could be surveyed on the extent to which they think their surrounding greenspace is attractive or not (e.g. not at all to strongly agree). This would provide a greater range of dependent values to map the variance of tree properties to, which may in turn grant new and insights into what factors shape greenspace attractiveness. Efforts at this future work would be important, because it could reveal new insights that could brighten the future environmental outlook for Vancouver.

Section 5: Interactive panel of findings

To highlight the main findings in this investigation, an interactive panel was created. All of the charts in the panel are interactive (with the exception of the bottom left chart).

To examine the claims of the current investigation:

Concluding remarks

The present investigation aimed to understand how attractive urban greenspaces could be created. To this end, it examined how urban forests in Vancouver are related to perceptions of community greenspace attractiveness. The findings revealed that more attractive greenspaces tend to contain a greater number of larger trees, whereas less attractive greenspaces tend to contain a greater number of smaller trees. Nevertheless, further work needs to be done before these findings could be applied to the real world. Specifically, future work should aim to replicate the results of the study with greater methodological rigor. Doing so may be important because it could reveal new insights that could brighten the future environmental outlook for Vancouver.

References